[quantization] Remove incorrect fp16 dynamic linear/relu op
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/32774
Test Plan: Imported from OSS
Differential Revision: D19624471
Pulled By: jamesr66a
fbshipit-source-id: eb6cb11fabf2ddd5edf345aff35b86b83c3af94c
